# Custom loss functions for XGBoost using PyTorch

**URL:** <https://forum.numer.ai/t/custom-loss-functions-for-xgboost-using-pytorch/960>\
**Category:** Data Science\
**Created:** [September 19, 2020, 9:28am UTC](https://forum.numer.ai/t/custom-loss-functions-for-xgboost-using-pytorch/960 "2020-09-19T09:28:20Z")\
**Posts on this page:** 1\
**Showing post:** 6

<div class="post-metadata">

**Author:** ![nrichers](https://avatars.discourse-cdn.com/v4/letter/n/8e7dd6/32.png) [@nrichers](https://forum.numer.ai/u/nrichers)\
**Post date:** [October 13, 2020, 4:55pm UTC](https://forum.numer.ai/t/custom-loss-functions-for-xgboost-using-pytorch/960/6 "2020-10-13T16:55:25Z")

</div>

I am aware of era-wise cv, I let `cv=3` for for simplicity.

When you say “ **write your own cross validation code** ”, you are suggesting to extend  
[BaseSearchCV](https://github.com/scikit-learn/scikit-learn/blob/0fb307bf3/sklearn/model_selection/_search.py#L400) class similarly you did [here](http://forum.numer.ai/t/era-wise-time-series-cross-validation/791)?

```auto
class TimeSeriesSplitGroups(_BaseKFold)

```

Sorry about the confusion with xgb docs…

---

_[View the full topic](https://forum.numer.ai/t/custom-loss-functions-for-xgboost-using-pytorch/960)._